Summary
This release includes updates to the source code and documentation for a better understanding of the model's implementation. It specifically adds comprehensive notebooks for computational performance benchmarking and multi-study correction analysis.
📄 New Notebooks
- Performance Benchmarking and Fine-tuning example: Added
docs/tutorial/report-performance.ipynb.- Example using a simulated dataset on fine-tuning the loss weights with a small subset of the training data.
- Provides detailed evaluation of wall-clock running times and peak GPU memory usage.
- Multi-studies Case: Added
docs/tutorial/demo-parkinson.ipynb.- Demonstrates model performance on a composite of 9 studies dataset (merging multiple test cases).
- Validates the preservation of biological signals without adding biases in high-heterogeneity settings.
- Inflammatory Bowel Disease Case Study: Added
docs/tutorial/tutorial-ibd.ipynb.- Added the implementation of the Inflammatory Bowel Disease (IBD) case study as a tutorial notebook.
- Dealing with outlier training runs for analysis of problematic taxonomic groups.
🛠 Source Changes
- Fixed a bug that wasn't allowing the use of the
VMMprior configuration atmetaABaCo().